Proceedings of ASME 2016 International Manufacturing Science and Engineering ConferenceMSEC 2016

June 27 – July 1, 2016, Blacksburg, VA, USA

MSEC2016-8629

AUTOMATING ASSET KNOWLEDGE WITH MTCONNECT

Sid Venkatesh, Sidney Ly and Martin ManningBoeing Company

Seattle, WA

John Michaloski and Fred ProctorNational Institute of Standards

and TechnologyGaithersburg, Maryland, USA

ABSTRACTIn order to maximize assets, manufacturers should use real-

time knowledge garnered from ongoing and continuous collec-tion and evaluation of factory-floor machine status data. In dis-crete parts manufacturing, factory machine monitoring has beendifficult, due primarily to closed, proprietary automation equip-ment that make integration difficult. Recently, there has been apush in applying the data acquisition concepts of MTConnect tothe real-time acquisition of machine status data. MTConnect isan open, free specification aimed at overcoming the “Islands ofAutomation” dilemma on the shop floor. With automated assetanalysis, manufacturers can improve production to become lean,efficient, and effective. The focus of this paper will be on thedeployment of MTConnect to collect real-time machine status toautomate asset management. In addition, we will leverage theISO 22400 standard, which defines an asset and quantifies as-set performance metrics. In conjunction with these goals, thedeployment of MTConnect in a large aerospace manufacturingfacility will be studied with emphasis on asset management andunderstanding the impact of machine Overall Equipment Effec-tiveness (OEE) on manufacturing.

KeywordsMTConnect, asset management, Overall Equipment Effec-

tiveness (OEE), manufacturing, Computerized Numerical Con-trol (CNC), network

NomenclatureAGFM American Gesellschaft fr Fertigungstechnik und

MaschinenbauAGV Automated Guided VehicleAPI Application Programming InterfaceCMM Coordinate Measuring MachineCNC Computer Numerical ControlCTLM Contour Tape Laying MachinesDCOM Distributed Component Object ModelDST Dorries Scharmann TechnologieEDM Electro Discharge MachiningFocas Fanuc OpenFactory CNC API SpecificationsHSSB High Speed Serial BusHTML Hypertext Markup LanguageHTTP Hypertext Transfer ProtocolKPI Key Performance IndicatorMTC Manufacturing Technology ConnectOEE Overall Equipment EffectivenessOEM Original Equipment ManufacturerPC Personal ComputerSCM Service Control ManagerSHDR Simple Hierarchical Data RepresentationSPC Statistical Process ControlURL Uniform Resource LocatorXML eXtensible Markup LanguageXSD XML Schema Definition

1 INTRODUCTIONProduction knowledge consists of understanding, organiz-

ing, and managing the machines, processes, and the tasks to beperformed in a manufacturing facility. The focus of this paper

1

will be on managing the machines, commonly referred to as as-sets. In an industrial context, asset management maximizes theperformance of production resources for achieving manufactur-ing objectives – producing products faster, cheaper, and better.For shop floor equipment, this means having a clear understand-ing of how machines operate and how to improve manufacturing.Informative and timely asset management can be used to accu-rately assess manufacturing operation and to make adjustmentsto meet shifting manufacturing conditions. Example manufac-turing roles for asset management include:

• accumulating resource knowledge for calculating account-ing functions such as the actual machining cost of a part forbidding, and determining profits [1–4],• identifying production bottlenecks [5–8],• building up machine histories in order to perform predictive

maintenance [9–12],• incorporating equipment and process knowledge dynami-

cally on a machine [13–15],• recognizing excessively high asset utilization as a prerequi-

site to determining procurement needs [16–18].

Although timely machine status feedback during factory op-eration is not a complex concept, the collection and dissemi-nation of the necessary status data in a timely and integratedmanner has been a challenge within the discrete parts industries.The breadth of machines in the discrete industries is extensive,from additive and subtractive machine tools, robots, and auto-mated guided vehicles (AGV). Plus, vendors, Original Equip-ment Manufacturers (OEMs), and end-users in the discrete in-dustries have all been reluctant to adopt a global integrationsolution and instead prefer a proprietary approach. Clearly is-sues stem from the over–abundance of industrial standards fromwhich to choose [19–26]. MTConnect is a standard to addressmany of these shortcomings, and is gaining traction in the dis-crete parts industry [27–29].

This paper discusses the automated collection and analysisof real-time machine status data based on the integration of MT-Connect and machine status reporting. The second section givesa brief overview of MTConnect and the machine tool informationmodel used for status reporting. The third section gives a back-ground on ISO 22400 and its model of asset management. Thissection will include a mapping of MTConnect machine statusinto ISO 22400 asset management concepts. The fourth sectiondescribes a case study of asset management using MTConnecton the shop floor at a large aerospace factory. The final sectioncontains a discussion on the benefits of machine tool status datain terms of support asset management as well as the problemsencountered developing automated machine status feedback andthe future work envisioned in this area.

2 MTConnect OverviewIn order to reduce costs, increase interoperability, and max-

imize enterprise-level integration, the MTConnect specificationhas been developed for the discrete manufacturing industry [30].Although aimed at solving the “Islands of Automation” prob-lem prevalent in the discrete manufacturing industries, MTCon-nect is flexible and can be easily adapted to other asset manage-ment or other manufacturing applications [31]. MTConnect isa specification based upon prevalent Web technology includingeXtensible Markup Language (XML) [32] and Hypertext Trans-fer Protocol (HTTP) [33]. Using this prevailing technology andproviding free software development kits minimize the technicaland economic barriers to MTConnect adoption.

Figure 1 shows the basic elements in an MTConnect solu-tion. MTConnect “Agent” is a software process that acts as abridge between a MTConnect “Device” and a Client Applica-tion. An MTConnect “Device” is a piece of equipment, like aCNC machining center or robot, organized as a set of compo-nents that provide data. MTConnect defines Events, Samples,Conditions, and Asset data items, whose XML format is all rigor-ously specified by XML Schema Definition (XSD) [34] schemas.Optionally, an MTConnect “Adapter” can be installed natively onthe machine, which is a process that provides a communicationlink between a device and the agent. Agents can have a special-ized remote Adapter or embedded “Backend” to communicateto the Device, (e.g., Simple Hierarchical Data Representation orSHDR [35] or OPC [36]).

The MTConnect standard defines XML schemas in order toexchange standard XML information. MTConnect defines theXSD content for Devices, Streams, Assets, or Errors necessaryfor retrieving factory device data, where:

• The MTConnect Devices XSD is an Information Model thatdescribes each device and its data items available.• The MTConnect Streams XSD is an Information Model that

describes a time series of data items specified in the DevicesXML including samples, events, and conditions.• The MTConnect Error XSD defines the XML to describe

one or more errors that occurred in processing an HTTP re-quest to an MTConnect Agent.• The MTConnect Asset XSD defines the XML pertaining to

a machine tool asset, which is not a direct component ofthe machine and can be relocated to another device duringits lifetime. The concept of MTConnect Asset refers to com-munication of resource asset knowledge and not the physicalresource discussed in this paper. MTConnect Asset exam-ples include tooling, parts, and fixtures.

Currently, MTConnect Agent supports four main types ofrequests:

• Probe request – response describes the devices whose datais being reported.

2

FIGURE 1: MTConnect Solution Overview

• Current request – retrieves the values of the devices dataitems at the point the request is received.• Sample request – retrieves a list of past and/or current values

for one or more data items.• Asset request – retrieves data describing the state of an asset.

Within an asset stream, there exists the ability to embed thirdparty developed standards, (e.g., ISO 13399 [37]) within theresponse.

HTTP is the protocol used by MTConnect (as well as theWorld Wide Web) to define legal messages [38]. MTCon-nect also establishes what constitutes legal commands throughthe use of decorated Uniform Resource Locator (URL) suchas http://agent.MTConnect.org/probe. MTConnect follows therules of HTTP to fetch and transmit the requested MTConnectcommand, be it “probe”,“current”,“sample”, or “asset”.

In a “probe”, an MTConnect Device is modeled in XMLwhich conforms to the Device XSD. The “Device Data Model”provides the Device(s) description that the world will see, whichwill typically be a subset of the total possible data from a device.The MTConnect Device model is not hardwired; rather users as-semble an XML information model to match their device andtheir data requirements. Each MTConnect implementation usesa Device Data XML document to describe the data that will beconveyed from one or more devices. In effect, of the thousands ofdata items that may be available from a controller, MTConnectprovides an XML document that enumerates which data itemsare in fact available. For example, suppose an MTConnect useris interested in the parameter “following error” of a servo drive,then the user would have to see if the Device is configured to

supply “following error” data.MTConnect Stream and Device XML Document are simi-

lar to all XML documents in that they are a tree representation.At the root, “Devices” define one or more “Device” (e.g., ma-chine tool), which in turn defines a set of components, which inturn contain “Data Items”. Thus, an “MTConnect Device” is amachine organized as a set of components that provide data. Fig-ure 2 shows a simple MTConnect hierarchy. In this MTConnectexample, “cnc1” is composed of components: “power”, “con-troller”, and “axes”. Each component then has event or sam-ple Data Item definitions. In this example, the “axes” compo-nent has sample data items: Srpm, Xabs, Yabs, and Zabs. Incontrast, the “controller” component has two event data items:mode, and execution; and one sample data item: feedrate. Sam-ple tags (e.g., Xabs, Yabs, Zabs) exhibit numerical values asstrings. Some event tags have an enumeration string, e.g., themode event can be either: MANUAL, MANUAL DATA EN-TRY, AUTOMATIC.

Generally, MTConnect performance is low bandwidth (i.e.,1 to 2Hz), so that start/stop/program changes/alarms and othermachine specific events are easily identified at this sampling rate.Higher bandwidth techniques are available in MTConnect, butare out of scope for this document.

3 ASSET MANAGEMENTManufacturing companies create products by converting raw

materials, stock, or supplied goods into a finished product tosell. In general, manufacturing is complicated due to the parallelmachine operation, dynamic job arrival, multi-resource require-

3

FIGURE 2: MTConnect Device Hierarchy Example

ments, and general job precedence constraints. At its most basic,manufacturing is handled by a set of machines, with a varyingdegree of flexibility and control; a material handling system thatallows jobs to move between machines; and a set of computersfor command and control. Preferably, the set of computers is allfully integrated on an enterprise network.

Asset management quantifies the performance of productionresources in achieving manufacturing objectives. Asset manage-ment can be used to build up machine histories, which can beused to make informed business decisions, such as, capital avoid-ance (when to hold off on equipment purchases) or intelligentpurchasing (formal evaluation of a given CNC model’s actualperformance, not anecdotal based on estimated data). Asset man-agement can be used to store information on equipment withinprocesses for use in undertaking corrective actions. The patternsof equipment operation can be analyzed for bottlenecks or under-utilization and then used to improve the production process andreduce costs and improve product turnaround.

Depending on the individual or organization, possibly fromdifferent technical and geographic points of origin, assets mayhave a differing interpretation. Therefore, the consistent defini-tion of an asset must be in place that is universally understoodand adopted. Fortunately, the International Organization forStandardization (ISO) has established a manufacturing standard,ISO 22400 [39] “Automation systems and integration Key per-formance indicators (KPIs) for manufacturing operations man-agement” that defines the concept of an asset and quantifies theassociated performance metrics.

FIGURE 3: ISO 22400 Work Unit Formalism

Foremost, ISO 22400 offers consistent manufacturing con-cepts and terminology. Such that if KPIs are to be used in mul-tiple locations and are to be searched, shared, and analyzed, acommon vocabulary (as well as models) is a prerequisite. In ad-dition, unnecessary cost from mistranslation, misunderstanding,and misinterpretation is avoided. Thus, common terminologyand models are helpful in identifying and monitoring enterpriseneeds and outcomes by pooling data from multiple sources in asystematic method.

The ISO 22400 standard is presented according to high-levelISA 95 Manufacturing Operations Management (MOM) [40] in-formation categories – the machinery and equipment, the productmanufactured and its quality, the manufacturing personnel, theinventory, and other related manufacturing elements. AlthoughISO 22400 covers a complete production model, of interest tothis paper is the ISO 22400 formalism to model individual assetsor “Work Units” (i.e., ISA 88 terminology for resources or ma-chines [41]) on the shop floor. ISO 22400 includes factors suchas costs, quality, time, flexibility, environmental and social is-sues, and energy efficiency many of which are important but outof scope for this paper.

ISO 22400 formally breaks down the Work Unit produc-tion model into planned activities and actual production. Fig-ure 3 shows the ISO 22400 formalism for “planned” and “ac-tual” Work Unit modeling used to assess asset performance.ISO 22400 states that a day is the planned maximum time avail-able for production and maintenance tasks, and a day depends onthe number of shifts. In ISO 22400, OEE is calculated by theequations below:

Availability = PBT/PDT

= PlannedBusyTime/PlannedProductionTime

E f f ectiveness = (PTU×PQ)/PDT

= (Production time per unit×Produced quantity)/PDT

4

Quality = PartCountGood/(PartCountGood +PartCountBad)

OEE = Availability×E f f ectiveness×Quality

In above equations, Availability, E f f ectiveness, Quality andOEE units are percent. The Availability determines how stronglythe capacity of the machine for the value-added functions re-lated to the planned availability is. Availability takes into ac-count planned time loss, e.g., meetings, coffee breaks, and main-tenance. E f f ectiveness is the measure for the efficacy of aprocess comparing target cycle time to the actual cycle time.E f f ectiveness is also called efficiency factor or performance.The Quality rate is the relationship of the proper quantity to theproduced quantity. Availability takes into account planned downtime loss, E f f ectiveness takes into account delays and speedloss, and Quality takes into account part loss.

Unfortunately, OEE can be rendered meaningless due to thelack of appropriate data. Specifically, OEE requires a qualitycomponent in its calculation, which in discrete parts productionis often impossible or can be difficult to determine since qualityassessment is typically done later in the manufacturing processand is disconnected from the machining process. In this caseOEE degrades into an asset utilization metric. For our analysis,we will assume quality is a meaningful component.

3.1 Mapping ISO 22400 to MTConnectISO 22400 distinguishes between performance data (such

as Work Unit busy, delay, down, queued, etc.) and KPI (e.g.,Work Unit OEE) [42]. It is the role of MTConnect to supplythe machine status data. However, ISO 22400 is geared towardasset management with production flow between machines. Inother words, ISO 22400 defines delay to include the concepts ofblocked, starved, and queued as well as faulted or idle. For thecase study that follows, a job shop is a more appropriate model ofmachine–part relationship and the concepts of blocked, starved,or queued are generally not relevant in a job shop part flow.

The terminology correspondence between ISO 22400 andMTConnect is not a perfect match, so it is best to clarify thedata terminology. For ISO 22400, “Down” is the same con-cept as machine “Off” . Likewise the ISO 22400 concept of“Delay” incorporates the concepts “Idle” and “Faulted”. Longterm “Faulted” effects the OEE Availability of the machine, andwould be a loss due to unscheduled maintenance. This leads toan ISO 22400 state model of down, idle, faulted, or busy whileignoring the starved and blocked states. (Idle as represented bythe part queued state is a data parameter for a separate KPI.)

For MTConnect systems, the basic process to provide OEEperformance data is to interpret MTConnect state logic usingmode, execution, and other MTConnect tags to determine themachine status of busy, idle, faulted, and off. Although different,it is possible to map the MTConnect data into asset managementstate model. Table 1 shows the mapping of MTConnect status

data items into state formalism corresponding to the ISO 22400asset model. The MTConnect Data row (first row) details the ex-pected raw data available from the MTConnect Device. The firstcolumn abbreviations (e.g., APT, ASUT, ADET) correspond tothe ISO 22400 abbreviations from Figure 3.

TABLE 1: Mapping MTConnect State Data into Machine StatusData

Data Parameters

MTConnect Data Timestamp, Machine, Power, Mode, Execu-tion, Spindle, Conditions, Feed override

Parameters Data Mapping

APT ≡Busy

MTCprogram 6= /0 and MTCexecution = Active

ASUT ≡Setup

MTCmode ∈Manual

ADET ≡Delay

MTCprogram = /0 or MTCexecution =Stopped or MTCexecution = Interrupted orMTCmode = Manual or Faulted

Down ≡(Off)

MTCpower = O f f

Faulted ∃ MTCcondition = f aulted until∀MTCcondition 6= f aulted

Some of the OEE loss is not explicitly covered byISO 22400. For example, MTC f eedoverride<100 % implies thatthe operator is slowing down machining and increasing cycletime, either to avoid chatter or for some other reason. Thisnegatively impacts OEE E f f ectiveness ratio. Another exam-ple of OEE E f f ectiveness loss, that is not directly covered inISO 22400, is if the operator is performing first part dry runtesting, when spindle rpm = 0. Likewise the operator could beprobing the part with machine axes moving but again spindlerpm = 0.

4 CASE STUDYA case study was performed that investigates the MTCon-

nect status data requirements for an aerospace manufacturing fa-cility that produces a wide variety of airplane parts, (e.g., brack-ets, body joints, etc.). The Boeing Company Auburn facility is195000 square meters (2.1 million-square-foot) and is reportedlythe largest airplane parts plant in the world, making and stor-ing more than 200000 parts for commercial jetliners [43]. Partmaterials vary and include aluminum, stainless steel, titanium,and inconel. The facility can produce parts ranging from a few

5

ounces to over a ton, with dimensional control in the range of±0.0254 mm to ±0.00254 mm (±0.001 inch to ±0.0001 inch).The disparity of part requirements means that there are manytypes of manufacturing machines, including horizontal mills,vertical mills, routers, waterjet, and wire electro discharge ma-chining (EDM). For confidentiality, the actual performance datahas been normalized; however, the analysis is representative ofthe data that is frequently encountered in facilities such as theone described in this study.

The flow of parts through the facility is determined by aworkorder for each part(s). A process planner prepares a worko-rder for a part(s) that assigns resources, which incorporatesconstraints based on the asset configuration (e.g., 3 versus 5axis), surface finish (e.g., high speed machining versus normalmilling), machine horsepower, and feature tolerances, among amyriad of part requirements. At the same time the workorder isprepared, the corresponding part program based on the versionand revision of the part is uploaded to the program database.The workorder is eventually routed to a machine operator, andin preparation uses one of the designated asset types, gets theraw material, downloads the part program from the database, anddoes a set up according to the workorder. Should the workorderspecify that a large number of parts are to be made, a test tryout is done on one part to insure correctness of the plan. Oftenafter the test part is made, the part is inspected, and when thepart meets or exceeds the quality requirements, the remainder ofthe parts are machined according to the workorder. This processis repeated until the test part satisfies the quality requirements.Overall, the flow of parts through the facility resembles a jobshop, as opposed to a production line.

The Boeing Auburn plant has been an early adopter of MT-Connect and has extensive MTConnect connectivity throughoutthe factory. The plant use of MTConnect, although not 100 %,encompasses a wide variety of machines and vendors: Mazak,Jomach, Northwood, DST, ATA Group, American Gesellschaftfr Fertigungstechnik und Maschinenbau (AGFM), and Nikon,as well as a variety of controllers: Original Equipment Man-ufacturer (OEM), Fanuc, Mitsubishi, and Siemens controllers.Most MTConnect Agents were installed as Windows Servicesthat ran as a 24/7 operation. Since the MTConnect agents were aservice and not application “exes”, only the Windows ServiceControl Manager (SCM) can start/stop the programs. When-ever possible, native MTConnect solutions from the CNC ven-dor were preferred. However, this is not always possible, socustom MTConnect solutions were developed. Fortunately, MT-Connect provides open source software solutions for the Agentand a wide range of Adapters, which can be found at https://github.com/MTConnect, and were used extensively.

Many of the implementations used an embedded MTCon-nect Adapter in the controller to perform the communication be-tween a device and the agent (e.g., SHDR, Fanuc High SpeedSerial Bus (HSSB) FOCAS). MTConnect also supported special-

(a) Mazak

(b) Siemens 840D Powerline

(c) Fanuc Focas HSSB

(d) Fanuc Focas Ethernet

(e) Logfile

FIGURE 4: MTConnect Case Study Solutions

ized “Backends” in the Agents to communicate to the device viasome other remote access protocol, (e.g., OPC, Fanuc EthernetFOCAS, Logfile). Below is a snapshot of some of the MTCon-nect solution strategies employed.

• Mazak provides native MTConnect support, supplying aSHDR Adapter and MTConnect Agent. Installation of theMTConnect components were handled by a Mazak servicerepresentative. Figure 4a shows the Mazak controller emit-ted SHDR data to the MTConnect Agent, which translatedthe data updates into XML.• Dorries Scharmann Technologie (DST), Jomach, and Echos-

peed CNC machines, use a Siemens 840D controller. Theolder Siemens 840D (i.e., Powerline models) support OPCData Access “Classic” [44]. Figure 4b shows two-way OPCcommunication using customized MTConnect Agent soft-ware with an OPC adapter backend. Heightened cyberse-curity precautions make the use of traditional OPC withDistributed Component Object Model (DCOM) connectionvery problematic.• Northwood, Cincinnati, Heian Router, and numerous con-

tour tape laying machines (CTLM) used a Fanuc controller.There were many different Fanuc iSeries controller modelswith subtle differences in functionality. Fanuc machines re-quired the FANUC OpenFactory CNC API Specifications

6

(Focas) library to communicate to the controller. Focas pro-vides a DLL with API to manage and query the state of theCNC machine. Fanuc supplies two communication methodsusing Focas: High Speed Serial Bus (HSSB) and Ethernet.Figure 4c shows the HSSB Focas solution, where one FocasSHDR Adapter is installed on each CNC and then an MT-Connect Agent reads the Fanuc CNC data items using theSHDR protocol. Figure 4d shows the Ethernet Focas solu-tion, which allows MTConnect to remotely access status ofmultiple controllers.• For ATA Group, AGFM, and Nikon Coordinate Measuring

Machines, status data was obtained from the controller logfile. Figure 4e shows a Windows networked file sharing ap-proach used in conjunction with controller log file monitor-ing. The networked file sharing approach is an easy deploy-ment method as the controllers do not require any specialsoftware interface, and need not be aware of the MTCon-nect monitoring. Cybersecurity protection prevented tradi-tional file access as even file sharing from the MTConnectservice across the network needed logon authentication.

Cybersecurity and safety protection of the machines and hu-mans are major concerns. Two cybersecurity schemes were inplace. One scheme has a dual Ethernet solution where the MT-Connect Agent runs on a front end Personal Computer (PC) andtalks to the asset through a local network connection. The secondscheme uses a hardware firewall on the machine to block any net-work traffic, except from the computer running an MTConnectAgent. In each case the goal is to isolate the machine from thecorporate Intranet but still allow connectivity by MTConnect.

FIGURE 5: Dashboard Client Application Architecture

A stack light dashboard client application monitored enter-prise as well as the case study factory machine status in real-time.Figure 5 shows the dashboard architecture, which is a central-ized, Web-based software application that collects machine sta-tus data from a large collection of MTConnect agents scatteredthroughout the enterprise, not just at the Auburn facility. Allthe MTConnect agent front-ends were nearly identical, while the

dispersed assets themselves were wide-ranging with varied func-tionality. All machine status data was gathered using MTConnectusing the Intranet as the communication backbone. Managerscan visually see which machines are up as well milling and getan intuitive feel for factory performance. With some historicallogging of the factory dashboard performance and drilling downon the actual use of a set of machines, one can perform assetmanagement with real data.

Previous to MTConnect networking of assets, machine sta-tus monitoring would have to be done manually. Manual datacollection is error prone and sporadic. Moreover, with an auto-mated process, operators are able to spend less time on non-valueadded reporting activities and more on productivity-orientedtasks. However, this automated approach must be easy to in-tegrate or the benefits will never materialize.

The goal for Auburn and other world class discrete manu-facturers is to achieve an 85 % OEE. Numbers higher than 85 %could indicate a bottleneck. When MTConnect asset manage-ment is used in conjunction with real time production log files,actionable knowledge is garnered where bottlenecks can be de-termined so that resources can be directed to mitigating the prob-lem. Numbers lower than 85 % indicate lost productivity. A sim-ple example illustrates the lost revenue that is identifiable withreal OEE values [45]. To determine a part BuildTime given theasset OEE:

BuildTime = Idea Cycle Time/OEE

where Ideal Cycle Time is the time required to produce the prod-uct if the asset was producing at 100 % of planned OEE capacity.Assuming an Ideal Cycle Time 270 minutes leads to the equa-tions:

BuildTime = 270/.85

= 317 minutes

BuildTime = 270/.50

= 540 minutes

If we assume a machine burn rate of $1000 per hour, a loss ofapproximately 2 hours equates to $2000.

OEE done with MTConnect asset management also helpsprovide hard numbers when assessing capital expenditures. Anaccurate OEE can determine asset capacity when combined withactual customer demand for product based upon planned produc-tion time (PDT). Capacity will vary based upon the utilization ofthe asset which changes based on the PDT or Planned ProductionTime (i.e., number of shifts per day, number of days per week,breaks, holidays, etc.). Assuming a baseline PDT capacity of 3shifts/day reduced by breaks yielding 20.4 hours per day totalthat will be further limited by the OEE reflected in the Build-Time:

CurrentCapacity (CC) = PDT/BuildTime

7

.85×OEE = 1224 min/317 min

= 3.8 parts/day

.50×OEE = 1224 min/540 min

= 2.2 parts/day

Assuming a 30 day month and the customer wants 90 parts amonth, then if the OEE is 0.50 % additional assets will be re-quired to satisfy the customer needs. Likewise, using a similarapproach, an anticipated increase in part demand can be strate-gically planned to see if current capacity coverage is sufficient.One of the primary benefits of quantifying asset performance iscost avoidance.

5 DiscussionAsset management is the monitoring of machine operation

to quantify performance for manufacturing. The key goals ofasset management include trouble-free integration within the en-terprise, collection and management of asset OEE information,real-time monitoring for preventive and reactive maintenance,and introspective process monitoring for adaptive control and er-ror compensation. To attain a goal of automated asset manage-ment, MTConnect has been shown as a standard and facilitatedan automated way to integrate discrete factory floor machinesinto the enterprise.

This paper detailed the use of MTConnect at a largeaerospace facility. MTConnect is an open factory communica-tion standard that leverages the Internet and uses XML for datarepresentation. Reliance on dominant and standard technologymade integration easier. Deployment did include custom ap-proaches to integrating assets. These custom approaches lever-aged MTConnect open source solutions available for many con-troller models. In addition, third party integrators provide afford-able MTConnect solutions for a number of controllers. In spiteof the benefits, CNC vendor support for MTConnect varied, socomplete integration of MTConnect on all factory machines re-mains challenging.

Once integrated, asset management can be performed bycollecting machine status data. Traditionally, manual recordingof activities is used to assess productivity. The use of real-timemachine status and recorded historical activity are important de-velopments in the quest for improving manufacturing. Usingreal-time and historical data, one can verify that the machine uti-lization is running as expected and if not, actionable procedurescan be identified and undertaken.

One final caveat is in order. Measured OEE is necessary, butnot sufficient, in order to enact production improvements. Un-derstanding the type and severity of delays within production isrequired to remediate process problems and improve OEE. MT-Connect can offer insight into some of the delays associated with

an asset. Such problems can include the under-performing op-erator, difficult setup, or chatter among numerous potential trou-bles, which can require further examination to truly understandthe root cause. Clearly hard real data can be beneficial in under-standing the manufacturing process. Informed management canmake better decisions based on facts not guesses, which leads tobetter manufacturing.

In the discrete part industry, integrating enterprise assetsthrough their automated data collection is a daunting task. Au-tomated factory data collection using MTConnect need not becontained to simple asset management. Expanding the scope toinclude advanced asset techniques, such as prognosis and condi-tion based health monitoring, are possible given improved datacollection. However, just collecting the data may not be suffi-cient, as a results-driven, automated analysis of an asset is im-perative to taming the potential voluminous windfall of data. Inaddition, even simple asset management could benefit from auto-mated analysis and decision making, as automated and integrateddata collection in itself does not make an intelligent system.

DisclaimerCommercial equipment and software, many of which are

either registered or trademarked, are identified in order to ade-quately specify certain procedures. In no case does such identifi-cation imply recommendation or endorsement by Boeing or theNational Institute of Standards and Technology, nor does it im-ply that the materials or equipment identified are necessarily thebest available for the purpose.

REFERENCES[1] Stockton, D., Khalil, R., and Mukhongo, M., 2013. “Cost

model development using virtual manufacturing and datamining: Part I - methodology development”. InternationalJournal of Advanced Manufacturing Technology, 66(5-8),pp. 741 – 749.

[2] Ben-Arieh, D., and Lavelle, J. P., 2000. “Manufacturingcost estimation: Applications and methods”. Journal ofEngineering Valuation and Cost Analysis, 3(1), pp. 43 –55.

[3] Sormaz, D., Gannon, P., and Pulugurta, S., 2013. “Method-ology for feature modeling and cost estimation of large castparts”. In IIE Annual Conference and Expo 2013, pp. 2349– 2356.

[4] Jung, J.-Y., 2002. “Manufacturing cost estimation for ma-chined parts based on manufacturing features”. Journal ofIntelligent Manufacturing, 13(4), pp. 227 – 238.

[5] Dewa, M., and Chidzuu, L., 2013. “Managing bottle-necks in manual automobile assembly systems using dis-crete event simulation”. South African Journal of IndustrialEngineering, 24(2), pp. 155 – 166.

8

[6] Lenort, R., Klepek, R., Wicher, P., and Besta, P., 2013. “Amethodology for determining and controlling the buffersbefore floating bottlenecks in heavy machinery produc-tion”. Metalurgija, 52(3), pp. 391 – 394.

[7] White, T., Sengupta, S., and Vantil, R. P., 2012. “A new wayto find bottlenecks”. Industrial Engineer, 44(11), pp. 45 –49.

[8] Lin, L., Qing, C., Guoxian, X., and Ambani, S., 2011.“Throughput bottleneck prediction of manufacturing sys-tems using time series analysis”. Journal of ManufacturingScience & Engineering, 133(2), pp. 021015–1 – 021015–8.

[9] Araiza, M. L., 2004. “A formal framework for predictivemaintenance”. In AUTOTESTCON (Proceedings), pp. 489– 495.

[10] Susto, G. A., Wan, J., Pampuri, S., Zanon, M., Johnston,A. B., O’Hara, P. G., and McLoone, S., 2014. “An adap-tive machine learning decision system for flexible predic-tive maintenance”. In IEEE International Conference onAutomation Science and Engineering, Vol. 2014-January,pp. 806 – 811.

[11] Ciocoiu, L., Hubbard, E.-M., and Siemieniuch, C. E., 2015.“Implementation of remote condition monitoring systemfor predictive maintenance: An organisational challenge”.In Contemporary Ergonomics and Human Factors 2015,pp. 449 – 453.

[12] Sipos, R., Fradkin, D., Moerchen, F., and Wang, Z., 2014.“Log-based predictive maintenance”. In Proceedings of theACM SIGKDD International Conference on KnowledgeDiscovery and Data Mining, pp. 1867 – 1876.

[13] Deshayes, L., Welsch, L., Donmez, A., Ivester, R., Gilsinn,D., Rhorer, R., Whitenton, E., and Potra, F., 2006. “Smartmachining systems: issues and research trends”. In Inno-vation in Life Cycle Engineering and Sustainable Develop-ment, D. Brissaud, S. Tichkiewitch, and P. Zwolinski, eds.Springer Netherlands, pp. 363–380.

[14] Suh, S., Kang, S., Chung, D., and Stroud, I., 2008. Theoryand Design of CNC Systems. Springer Series in AdvancedManufacturing. Springer London.

[15] Grabowski, R., Denkena, B., and Kohler, J., 2014. “Predic-tion of process forces and stability of end mills with com-plex geometries”. Procedia CIRP, 14, pp. 119–124.

[16] Leitner, K.-H., 2011. “The effect of intellectual capital onproduct innovativeness in SMEs”. International Journal ofTechnology Management, 53(1), pp. 1–17.

[17] Chen, Y.-S., and Chen, B.-Y., 2009. “Using data envel-opment analysis (DEA) to evaluate the operational perfor-mance of the wafer fabrication industry in taiwan”. Journalof Manufacturing Technology Management, 20(4), pp. 475– 488.

[18] South African Institute of Electrical Engineers, 2003. “ITin manufacturing”. Elektron, 20(4), pp. 48 – 49.

[19] INTERNATIONAL ORGANIZATION FOR STANDARDIZA-

TION (ISO), 2003. ISO 11898-1:2003 Road vehicles –Controller area network (CAN) – Part 1: Data link layerand physical signalling.

[20] INTERNATIONAL ELECTROTECHNICAL COMMISSION(IEC), 2003. IEC 61158 Type 2 - EtherNet/IP: Communi-cation Profile Family 2.

[21] INTERNATIONAL ELECTROTECHNICAL COMMISSION(IEC), 2014. IEC 61158 Type 10 - PROFINET: Communi-cation Profile Family 3.

[22] INTERNATIONAL ELECTROTECHNICAL COMMISSION(IEC), 2014. IEC 61158 Type 23 - CC-Link IE: Communi-cation Profile Family 8.

[23] INTERNATIONAL ELECTROTECHNICAL COMMISSION(IEC), 2014. IEC 61158 Type 12 - EtherCAT: Communi-cation Profile Family 12.

[24] INTERNATIONAL ELECTROTECHNICAL COMMISSION(IEC), 2014. IEC 61158 Type 13 - Powerlink: Communi-cation Profile Family 13.

[25] INTERNATIONAL ELECTROTECHNICAL COMMISSION(IEC), 2014. IEC 61158 Type 15 - Modbus-TCP:Communication Profile Family 15.

[26] CZECH NATIONAL STANDARD EUROPEAN STANDARD,2002. CSN EN 50325 - Industrial communications subsys-tem based on ISO 11898 (CAN) for controller-device inter-faces.

[27] Modern Machine Shop. Data-driven manufacturing movesahead at mazak. www.mmsonline.com/articles/

data-driven-manufacturing-moves-ahead-at-

mazak. [Online; accessed 20-Nov-2015].[28] Okuma America. MTConnect. www.okuma.com/

mtconnect. [Online; accessed 20-Nov-2015].[29] Mazak. MTConnect. www.mazakusa.com/machines/

mtconnect/. [Online; accessed 20-Nov-2015].[30] MTConnect Institute. MTConnect Specifications. www.

mtconnect.org/standard. [Online; accessed 25-Sept-2015].

[31] Edstrom, D., and Leonard, S., 2013. MTConnect: How andWhy a Royalty-Free and Open-Source Standard Is Revolu-tionizing the Business and Technology of Manufacturing:to Measure Is to Know. WordStyle LLC.

[32] World Wide Web Consortium, 2006. Extensible markuplanguage (XML) 1.0 (fourth edition).

[33] The Internet Society, 1999. Hypertext transfer protocol -http/1.1.

[34] World Wide Web Consortium, 2004. XML schema part0: Primer second edition. www.w3.org/TR/2004/REC-

xmlschema-0-20041028. [Online; accessed 20-Nov-2015].

[35] Sobel, W. MTConnect Standard SHDR Protocol Compan-ion Specification. https://github.com/mtconnect/

adapter/blob/master/MTC_SHDR.docx. [Online; ac-cessed 25-Sept-2015].

9

[36] OPC Foundation. OPC. https://opcfoundation.org.[Online; accessed 25-Sept-2015].

[37] INTERNATIONAL ORGANIZATION FOR STANDARDIZA-TION, 2014. ISO 13399, Cutting tool data representationand exchange. Geneva, Switzerland.

[38] MTConnect User Portal. MTConnect HTTP Primer.mtcup.org/wiki/HTTP_Primer. [Online; accessed 25-Sept-2015].

[39] INTERNATIONAL ORGANIZATION FOR STANDARDIZA-TION (ISO), 2014. ISO 22400-1:2014: Automationsystems and integration Key Performance Indicators(KPIs) for manufacturing operations management Part 1:Overview, concepts and terminology. Geneva, Switzerland.

[40] ISA, 2005. ANSI/ISA-95.00.03-2005, Enterprise-ControlSystem Integration, Part 3: Models of Manufacturing Oper-ations Management. International Society of Automation.

[41] ISA, 2010. ANSI/ISA-88.00.01-2010 Batch Control Part 1:Models and Terminology. International Society of Automa-tion.

[42] INTERNATIONAL ORGANIZATION FOR STANDARDIZA-TION (ISO), 2014. ISO 22400-2:2014: Automation sys-tems and integration Key Performance Indicators (KPIs)for manufacturing operations management Part 2: Defini-tions and descriptions. Geneva, Switzerland.

[43] KOMO Television. Boeing’s Auburn plant declaredsafe. www.komonews.com/news/boeing/27937709.

html. [Online; accessed 20-Nov-2015].[44] OPC Foundation. OPC Data Access Clas-

sic. opcfoundation.org/developer-tools/

specifications-classic/data-access/. [Online;accessed 20-Nov-2015].

[45] Salter, J., 1998. Overall equipment effectiveness (OEE).Tech. rep., Boeing Company.

10